Sinclair BASIC

Sinclair BASIC is a dialect of the programming language BASIC used in the 8-bit home computers from Sinclair Research, Timex Sinclair and Amstrad. The Sinclair BASIC interpreter was written by Nine Tiles Networks Ltd.

Designed to run in only 1 kB of RAM, the system makes a number of decisions to lower memory usage. This led to one of Sinclair BASIC's most notable features, that the keywords were entered using single keystrokes; each of the possible keywords was mapped to a key on the keyboard, when pressed, the token would be placed into memory while the entire keyword was printed out on-screen. This made code entry easier whilst simplifying the parser.

The original ZX80 version supported only integer mathematics, which partially made up for some of the memory-saving design notes which had negative impact on performance. When the system was ported to the ZX81 in 1981, a full floating point implementation was added. This version was very slow, among the slowest BASICs on the market at the time, but given the limited capabilities of the machine, this was not a serious concern.

Performance became a more serious issue with the release of the ZX Spectrum in 1983, which ran too slowly to make full use of the machine's new features. This led to an entirely new BASIC for the following Sinclair QL, as well as a number of 3rd-party BASICs for the Spectrum and its various clones. The original version continued to be modified and ported in the post-Sinclair era.

History
Clive Sinclair initially met with John Grant, the owner of Nine Tiles, in April 1979 to discuss a BASIC for Sinclair's new computer concept. Sinclair was inspired to make a new machine after watching his son enjoy their TRS-80, but that machine's £500 price tag appeared to be a serious limit on its popularity. He wanted a new kit that would expand on their previous MK14 and feature a built-in BASIC at the target price of £79.95. To meet this price point, the machine would ship with only 1 kB of RAM and 4 kB of ROM. Grant suggested using the Forth language instead, but the budget precluded this. Grant wrote the BASIC interpreter between June and July 1979, but the code initially came in at 5 kB and he spent the next month trimming it down. It was initially an incomplete implementation of the 1978 American National Standards Institute (ANSI) Minimal BASIC standard with integer arithmetic only, termed 4K BASIC.

Even before the ZX80 was introduced in February 1980, the constant downward price-pressure in the industry was allowing the already inexpensive design to be further reduced in complexity and cost. In particular, many of the separate circuits in the ZX80 were re-implemented in a single uncommitted logic array from Ferranti, which allowed the price to be reduced to only £49.95 while increasing the size of the ROM to 8 kB. This work was assigned to Steve Vickers, who joined Nine Tiles in January 1980. Whilst Grant worked on the code interfacing with hardware, Vickers used the larger space to introduce floating-point arithmetic and a suite of trig functions, which were expected of any BASIC from that era, producing 8K BASIC. The initial version did not support the ZX Printer and had a bug in its square root function. Nine Tiles provided a new version to address these, but Sinclair was slow to include the new version in the ROMs. The new ROMs were eventually offered to owners of the earlier ZX80 as well.

When Sinclair lost the contest to build the BBC Computer, he moved ahead with plans to produce a low-cost colour-capable machine that emerged as the ZX Spectrum of April 1982. ROM space would once again be increased, this time to 16 kB. In keeping with his philosophy of making systems for the lowest possible expenditure, Sinclair wanted the absolute minimum changes to the existing 8K BASIC. Although Nine Tiles felt that something much better would be needed for the new machine, the schedule would not allow it, and yet another expansion of the original code was produced. Due to the RAM also being increased, to 48 kB, this version was known as 48K BASIC and eventually 48 BASIC with the introduction of the ZX Spectrum 128 at which time the 16 KB Spectrum was no longer sold and most existing ones in use had been upgraded to 48 KB. The new version was available near the end of 1981, but it was "depressingly slow" and "snail like". Additionally, as no prototypes were available until the end of the year, it lacked support for the new line of peripherals Sinclair was planning.

In February 1982, Nine Tiles began to have disagreements with Sinclair over owed royalties for the various manuals that Nine Tiles had produced. Around the same time, Vickers and his Sinclair counterpart, Richard Altwasser, left their respective companies to start the Jupiter Ace project. When the Spectrum was launched the ROM was still not complete, and although Nine Tiles continued working on it until April 1982, by that point 75,000 Spectrums had already been sold and the project was cancelled. The missing functionality was later added by additional code in the ZX Interface 1. After Nine Tiles and Sinclair went their own ways, several new versions of 48 BASIC were created.

In 1983, as part of introducing the Spectrum to the US market as the TS2068, Timex modified it as T/S 2000 BASIC. The new version was incapable of running many Spectrum programs due to the memory location of machine's functions moving. A similar, but somewhat more compatible version, was introduced as part of the Spanish Spectrum 128 of 1985, 128 BASIC.

Amstrad purchased Sinclair Research in 1986. As Sinclair had never owned the copyright to the language, Amstrad had to arrange a new license with Nine Tiles. Several other versions also appeared in this period as various extensions and clones of the Spectrum were introduced. These included +3 BASIC, BASIC64 and Timex Extended Basic.

As of 2015, interpreters exist for modern operating systems, and older systems, that allow Sinclair Basic to be used easily.

Program editing
Like most home-computer BASICs, Sinclair BASIC is anchored around its line editor. When the machine is booted, it runs BASIC and displays an inverse video "K" at the bottom of the screen to indicate the entry point. When a line is entered and the key is pressed, it either runs immediately if it does not have a line number prefix, or clears the screen and performs the equivalent of a LIST command, placing a ">" cursor after the line number in the most recently entered line.

In contrast to most machines of the era, the editor does not allow freeform editing at any point on the screen. Instead, when the user presses EDIT, the current line of code is copied back to the bottom of the screen. The user can move horizontally through this line using the cursor keys and commits their changes by pressing NEW LINE again. In contrast, on machines like the Commodore 64 or Atari 8-bit computers, the up and down keys can be used to move among the lines in the program and edit them in-place.

The most notable feature of the editor is that keywords are entered using single keystrokes. For instance, on the ZX81, the P key on the keyboard would cause the entire keyword PRINT to be entered into the currently editing line. Once a keyword has been entered, the cursor changes to an "L" to indicate what follows will be interpreted as normal text. For instance, pressing P again at this point would enter a single letter "P".

Keys generally had two separate keywords assigned to them – one above it and one below. Pressing the key in "K" mode would enter the keyword above the key, like PRINT for. If the system was in "L" mode, one could return to "K" by pressing the shift key – the systems did not initially support lowercase text, so the shift was not otherwise needed. The keywords below the keys required a second keystroke, SHIFT, which put the editor into "function mode", changing the cursor to an "F". Entering common code often resulted in a significant number of keystrokes.

The system has the advantage of representing all multi-character keywords as a single character in memory, which was a significant savings in the early machines that shipped with only 1 kB of RAM. This single-character representation included multi-character items like <>. This has the added advantages of simplifying the runtime, as it can immediately determine whether a character in the source code is a keyword or text, and also means that keywords are never entered directly, meaning that one can, for instance, have a variable named "PRINT", as the system can determine that it is not the same as the keyword.

As the systems evolved and added new keywords, the entry system became increasingly difficult to use. 48 BASIC in the Spectrum required every key to host up to four keywords. Entering keywords was a time-consuming process of looking over the relatively small type on the keyboard for the appropriate key, and then correctly entering the multiple keystrokes needed to enter it properly. For instance, entering BEEP required one to type to access extended mode (later models include a separate  key), keeping  held down, and then and pressing. To improve the complex entry on the Spectrum, the keywords were colour-coded to indicate the required mode:
 * White: key only
 * Red on the key itself: plus the key
 * Green above the key: followed by the key
 * Red below the key: followed by  plus the key

This concept had run its course, and later machines running 128 BASIC (ZX Spectrum 128, +2, +3, +2A, and +2B) featured a more traditional editor where the user typed-in the keyword as individual characters, similar to other home computers of the era. This required a new tokenizer to convert the line into a similar internal format.

The resulting in-memory storage of the program was otherwise similar to Microsoft BASIC, in that only the keywords are presented as tokens, while non-keywords – like string and numeric constants and variable names –  are left in their original typed-in format. However, that typed-in format was not ASCII, but an internal character code that contains both printable characters and the keyword tokens. Although portions of the table, the capital letters A to Z, for instance, are in the same order as in ASCII, their characters correspond to different numeric values, i.e. their code points' offset from zero is different from ASCII's offset of 65. Other characters, like punctuation, may have been moved about even more. The ZX Spectrum character set is the most prominent example of such a character code. Related computers running Sinclair BASIC used similar variants, e.g. the ZX80 or ZX81 character sets. All of these different but related character sets included Sinclair BASIC tokens.

Data types
One uncommon feature of Sinclair BASIC is the way it stores variables in memory. Typically, interpreters use a fixed-size entry to hold data, making it easy to scan the variable table. Due to the extremely limited memory of the ZX series, any wasted space had to be avoided, and this led to the use of a variable-length format. The data types included numbers stored in a 5-byte values, strings with a length and then the characters, and arrays of both of those types. The data was stored in the table itself, which contrasts with most BASICs of the era, where strings and array entries were stored in a separate heap.

The first byte for a variable entry always held the type in the first three bits, and the first character of the name in the next five bits. As was the case in most microcomputer dialects,,  ,   and   were all different variables and could store different values. Most variables could only have a single-character name; the exception are numeric variables (not arrays), where an alternate format held the first character of the name in those same five bits, but was then followed by additional characters ending with one with its high-bit set. Long variable names were whitespace-independent, and case-insensitive in later versions, so  is the same as   referred to the same variable.

The downside to this approach is that scanning the table to look up the value of a variable reference is more complex. In addition to testing whether the name matches using the subset of the first byte, if the entry is not the one that is being looked for, the type has to be read from the upper three bits and then the next location of a variable in storage calculated using the type. For instance, if the program encounters the variable "A" and the table starts with the entry for "B$", it fails to match A with B, then reads the type to see it is a string, and then has to read the following length byte and skip forward by that amount of bytes to find the next entry in the table. To make this somewhat easier, arrays also stored a two-byte length, so the entire structure could be skipped over more easily.

A unique feature was the "short float", or integer type. Any numeric variable could store either type, the storage itself did not change and used 5 bytes in either case. Integers were indicated by setting the exponent byte to zero, while floating-point values were stored with an excess-128 format exponent. This meant that it could not store zero as a float and lost one possible exponent magnitude. It also did not use any less memory, as the values were still 5-byte in memory. The advantage to this format is performance; the math library included tests to look for the zero exponent, and if it was seen, it would not attempt to perform various operations on the remaining 3 bytes under certain conditions.

Keyword details
The ZX81 8K BASIC used the shorter forms,  ,   and  , whereas the Spectrum 48 BASIC used the longer forms  ,  ,   and. The ZX80 4K BASIC also used these longer forms but differed by using the spelling. The ZX81 8K BASIC was the only version to use,  ,   and. The ZX80 4K BASIC had the exclusive function ; it was equivalent to the string operator  in later versions.

Unique code points are assigned in the ZX80 character set, ZX81 character set and ZX Spectrum character set for each keyword or multi-character operator, i.e.,  ,  ,   (tokenized on the ZX81 only),   (replaced with   on the Spectrum). These are expanded by referencing a token table in ROM. Thus, a keyword uses one byte of memory only, a significant saving over traditional letter-by-letter storage. This also meant that the BASIC interpreter could quickly determine any command or function by evaluating one byte, and that the keywords need not be reserved words like in other BASIC dialects or other programming languages, e.g., it is allowed to define a variable named  and output its value with. This is also related to the syntax requirement that every line start with a command keyword, and pressing the one keypress for that command at the start of a line changes the editor from command mode to letter mode. Thus, variable assignment requires  (i.e.,   not only  ). This practice is also different from other BASIC dialects. Further, it meant that unlike other BASIC dialects, the interpreter needed no parentheses to identify functions;  was sufficient, no   needed (though the latter was allowed). The 4K BASIC ROM of the ZX80 had a short list of exceptions to this: the functions,  ,  ,  ,  ,  ,   and   did not have one-byte tokens but were typed in letter-by-letter and required the parentheses. They were listed as the INTEGRAL FUNCTIONS on a label above and to the right of the keyboard.

128 BASIC, present on ZX Spectrum 128, +2, +3, +2A, and +2B, stored keywords internally in one-byte code points, but used a conventional letter-by-letter BASIC input system. It also introduced two new commands:


 * , which operated the 128k models' General Instrument AY-3-8910 music chip
 * , which switched the 128k Spectrum into a 48k Spectrum compatibility mode

The original Spanish ZX Spectrum 128 included four additional BASIC editor commands in Spanish, one of which was undocumented:


 * (to edit a line number or invoke the full screen string editor)
 * (to renumber the program lines)
 * (to delete program lines)
 * (to set the column width of the RS-232 device, but undocumented as the code was broken)

Unlike the,   and   functions used in the ubiquitous Microsoft BASIC dialects for home computers, parts of strings in Sinclair BASIC are accessed by numeric range. For example, gives a substring starting with the 5th and ending with the 10th character of the variable. Thus, it is possible to replace the  and   commands by simply omitting the left or right array position respectively; for example  is equivalent to. Further,  alone is enough to replace.

Keywords
On the 16K/48K ZX Spectrum (48 BASIC), there are 88 keywords in Sinclair BASIC, denoting commands (of which there are 50), functions and logical operators (31), and other keywords (16, including 9 which are also commands or functions):

4K BASIC

 * 4K BASIC for ZX80 (so named for residing in 4 KiB read-only memory (ROM)), was developed by John Grant of Nine Tiles for the ZX80. It has integer-only arithmetic.
 * System Commands:
 * Control Statements:
 * Input/Output Statements:
 * Assignment Statement:
 * Other Statements:
 * Functions:

8K BASIC

 * 8K BASIC is the ZX81 BASIC (also available as an upgrade for the ZX80 ), updated with floating-point arithmetic by Steve Vickers, so named for residing in 8 KiB ROM.
 * Statements:
 * Functions:

48 BASIC

 * 48 BASIC is the BASIC for the original 16/48 KB RAM ZX Spectrum (and clones), with colour and more peripherals added by Steve Vickers and John Grant. It resides in 16 KB ROM and began to be called 48 BASIC with the introduction of the ZX Spectrum 128 at which time the 16 KB Spectrum was no longer sold and most existing ones in use had been upgraded to 48 KB.

128 BASIC

 * 128 BASIC is the BASIC for the ZX Spectrum 128. It offers extra commands and uses letter-by-letter input.
 * New commands:

+3 BASIC

 * +3 BASIC is the BASIC with disk support for the ZX Spectrum +3 and +2A.
 * New commands:

T/S 2000 BASIC

 * T/S 2000 BASIC is used on the Spectrum-compatible Timex Sinclair 2068 (T/S 2068) and adds the following six new keywords:
 * deletes BASIC program line ranges.
 * is a function that gives the amount of free RAM.  will show how much RAM is free.
 * is an error-handling function mostly used as  or.
 * can be used to reset the behaviour of . It was also intended to reset peripherals.
 * controls the AY-3-8192 sound chip.
 * is a function that gives the position of the internal joystick (Timex Sinclair 2090).

BASIC64

 * BASIC64 by Timex of Portugal, is a software extension to allow better Basic programming with the 512×192 and dual display areas graphic modes available only on Timex Sinclair computers. This extension adds commands and does a complete memory remap to avoid the system overwriting the extended screen memory area. Two versions exist due to different memory maps - a version for TC 2048 and a version for T/S 2068 and TC 2068.
 * Prints to a specific output channel.
 * Lists the program to a specific output channel.
 * Clears both display areas.
 * Sets ink colour for both display areas
 * Sets paper colour both display areas
 * Selects the high / normal resolution modes.
 * Plots a pixel and updates the drawing position.
 * Draws a line from the previous PLOT position, supporting arc drawing
 * Draws a circle or oval, depending on screen mode.

Timex Extended Basic

 * Timex Extended Basic by Timex of Portugal is used on the Timex Computer 3256, adding TEC - Timex Extended Commands commands supporting the AY-3-8912 sound chip, RS-232 network and the 512x192 pixel high resolution graphic mode.
 * RAM drive commands:
 * RS-232 commands:
 * AY-3-8912 commands:
 * 512x192 resolution commands:

Interpreters for the ZX Spectrum family
Several ZX Spectrum interpreters exist.
 * Beta BASIC by Dr. Andy Wright, was originally a BASIC extension, but became a full interpreter.
 * YS MegaBasic by Mike Leaman.
 * ZebraOS by Zebra Systems in New York, a cartridge version of T/S 2000 BASIC that used the 512×192 screen mode.
 * Sea Change ROM by Steve Vickers and Ian Logan, modified by Geoff Wearmouth, a replacement ROM with an enhanced Sinclair BASIC.
 * Gosh Wonderful by Geoff Wearmouth, a replacement ROM that fixes bugs and adds a tokenizer, stream lister, delete and renumber commands.
 * OpenSE BASIC (formerly SE BASIC) by Andrew Owen, a replacement ROM with bug fixes and many enhancements including ULAplus support, published as open source in 2011

Compilers for the ZX Spectrum family
Several ZX Spectrum compilers exist.
 * HiSoft COLT Compiler (a.k.a. HiSoft COLT Integer Compiler)
 * HiSoft BASIC (a.k.a. HiSoft BASIC Compiler), an integer and floating-point capable compiler
 * Laser Compiler
 * Softek 'IS' Integer Compiler (successor to Softek Integer Compiler )
 * Softek 'FP' Full Compiler
 * ZIP Compiler

Derivatives and successors for other computers

 * SuperBASIC, a much more advanced BASIC dialect introduced with the Sinclair QL personal computer, with some similarities to the earlier Sinclair BASICs
 * SAM Basic, the BASIC on the SAM Coupé, generally considered a ZX Spectrum clone
 * ROMU6 by Cesar and Juan Hernandez - MSX
 * Spectrum 48 by Whitby Computers - Commodore 64
 * Sparky eSinclair BASIC by Richard Kelsh, an operating system loosely based on ZX Spectrum BASIC - Zilog eZ80
 * Sinbas by Pavel Napravnik - DOS
 * Basic (and CheckBasic ) by Philip Kendall - Unix
 * BINSIC by Adrian McMenamin, a reimplementation in Groovy closely modelled on ZX81 BASIC - Java
 * BASin by Paul Dunn, a complete Sinclair BASIC integrated development environment (IDE) based on a ZX Spectrum emulator - Windows
 * SpecBAS (a.k.a. SpecOS) by Paul Dunn, an integrated development environment (IDE) providing an enhanced superset of Sinclair BASIC - Windows, Linux, Pandora, and Raspberry Pi
 * ZX-Basic Backwardly compatible, but enhanced compiler of Sinclair BASIC programs the ZX Spectrum, written in Python, freely available for Windows, Linux and Mac OS
 * ZX-Basicus by Juan-Antonio Fernández-Madrigal, a synthesizer, analyzer, optimizer, interpreter and debugger of Sinclair BASIC 48K for PCs, freely downloadable for Linux and Windows.